Introduction

[Flavio Coelho]

Hello,

I have just joined this list.

My name is Flávio Coelho and though I have interacted with some
members of this community through comments in my blog,
pyinsci.blogspot.com, I thought it was appropriate to introduce
myself.

I work with computational models of population biology, and I have
been involved with Bayesian Inferential techniques for many years as
they are IMHO, the only viable way to handle the kind of uncertainties
we have in complex dynamical models of living systems.

I have been following PyMC development for quite a while now and I
would like to congratulate all of you for the tremendous effort put
into PyMC to make it into what it is today. I'd also like to help, if
I can, to make it even better.

One area I may be able to help is in the parallelization of chains in
the MCMC samplers using the multiprocessing module, if that problem is
not solved yet.

cheers,

Flávio

[Anand Patil]

Hi Flávio,

Welcome! I'm very happy to see you here. PyMC stands to benefit from
your fresh perspective & experience with Bayesian computing in Python.

David Huard is the point-person on multiple CPU's. John Salvatier has
also been working on a multiple-chain jumping strategy, which may be
able to make use of multiple processors.

We're currently working on a maintenance release, 2.1, to accompany
the impending publication of our manual in Journal of Statistical
Software. That release has been informally feature frozen, but it will
hopefully provide a solid foundation for innovation in the future.

Have fun and let us know if you have questions or concerns.

Anand

> --
>
> You received this message because you are subscribed to the Google Groups "PyMC" group.
> To post to this group, send email to py...@googlegroups.com.
> To unsubscribe from this group, send email to pymc+uns...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/pymc?hl=en.
>
>

[Flavio Coelho]

Hi Anand,

thanks for the warm welcome!

It will probably take a while to familiarize myself with the codebase.
Then i'll start to play around with some parallelization techniques,
and see if I can make some of the samplers faster.
The Multiprocessing is great for taking advantage of multiple
pcu/cores in SMP systems without adding external dependencies.
It gets more complicated if you want to enable parallel processes in
a cluster of machines.

Congratulations on the publication of the manual! it'll certainly
attract a lot of extra attention to PyMC.

Flávio

--
Flávio Codeço Coelho